干骨计算机断层扫描成像的优化后处理参数:面向多种研究应用的建议
本研究以49根考古出土的干人类股骨为对象,由两名盲法评审员按五点李克特量表对四种卷积核与重建矩阵组合的CT图像质量评分,并以光学扫描仪模型为金标准比较三维重建质量。结果显示Br59+512组图像质量评分最高,Br40+1024组在横截面积和体积测量上与参考组无统计学显著差异;作者推荐Br59卷积核配合512矩阵用于最佳图像观察,Br40卷积核配合1024矩阵用于三维重建与形态学分析。
要点
- Br59+512组合的CT图像质量评分最高。
- Br40+1024组横截面积和体积测量与参考组无统计学显著差异。
- 作者推荐Br59+512用于图像观察,Br40+1024用于三维重建。
核心结论
- 锐利卷积核与更高的图像质量相关,而软卷积核有助于降低噪声并提升三维重建质量;更大的重建矩阵基于视觉评估未显著改善图像质量,但有助于获得更高质量的三维模型。
研究设计与样本
- 研究选取49根考古遗址出土的干人类股骨,由两名盲法评审员使用五点李克特量表对四种卷积核与重建矩阵组合的CT图像质量评分,并以光学扫描仪生成的模型为金标准比较四种CT组的三维重建质量。
诊断标准与关键数值
- 图像质量评分采用五点李克特量表,Br59+512组得分最高。
- 三维重建长度测量中,Br40+512组与Br40+1024组差异有统计学意义(P < 0.05)。
- 角度测量在所有组间均无显著差异。
- 横截面积和体积测量中,Br40+1024组与参考组无统计学显著差异。
临床应用与阅片要点
- 作者建议图像观察采用Br59卷积核配合512矩阵。
- 作者建议三维重建与形态学分析采用Br40卷积核配合1024矩阵。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVE: This study aimed to compare the performance of different convolution kernels and reconstruction matrices in enhancing computed tomography (CT) image quality and three-dimensional (3D) reconstruction quality. Based on these findings, recommendations for optimizing CT post-processing parameters are provided to meet various research needs.
METHODS: Forty-nine dry human femurs, excavated from archaeological sites, were selected for analysis. Two blinded reviewers scored the CT images from four convolution kernel and reconstruction matrix combinations using a five-point Likert scale for image quality. For 3D reconstruction quality, models generated by optical scanners served as the gold standard and were compared with four CT groups. Measurements included length, angle, cross-sectional area, and volume.
RESULTS: CT images processed with the Br59 + 512 group achieved the highest image quality scores. For 3D reconstruction quality, a significant difference in length measurements was observed between the Br40 + 512 and Br40 + 1024 groups (P < 0.05). Angle measurements showed no significant differences across all groups. In cross-sectional area and volume measurements, the Br40 + 1024 group showed no statistically significant difference from the reference group.
CONCLUSION: Sharp convolution kernels are associated with higher image quality, whereas soft kernels help reduce noise and enhance 3D reconstruction quality. Although larger reconstruction matrices do not significantly improve image quality based on visual assessment, they contribute to higher-quality 3D models. We recommend the Br59 kernel with a 512 matrix for optimal image observation and the Br40 kernel with a 1024 matrix for 3D reconstruction and morphological analyses.
引用
- Xing H, Zhu J, Zou R, et al. Optimized post-processing parameters for dry bone computed tomography imaging: Recommendations for diverse research applications. Eur J Radiol. 2025. PMID: 40849990. doi:10.1016/j.ejrad.2025.112376
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。